The article presents structural models of terminological phrases from the subject area “Welding” as the basis for creating automated tools to mark up the corpus of scientific and technical texts. The place of scientific and technical corpora in corpus linguistics and the prospects for their further research are outlined. The relevance of the research stems from the need to create corpora of scientific and technical texts in general and to provide tools for automatic detection of terms in particular. It is substantiated that the main problem in designing such corpora is the automatic markup of terminological phrases. The analysis of the current state of the term system of the subject area “Welding” has been carried out. The results of the analysis of two-, three-, four- and five-component terminological phrases of “Welding” and their structural models are presented and illustrated by examples. The necessity of listing all possible structural models of terminological combinations has been substantiated too. It has been established that the addition of a new component to the basic terminological combination most often occurs with introduction of one more postpositional at-tribute whose function is to add some specific feature to the basic meaning. The novelty of the study is seen in providing a theoretical approach for the formation of a database of structural models of terminological phrases which may be used as a core of a supersource database on the structure of the multicomponent scientific and technical terms. An approach to automatic markup of multicomponent terms is proposed too. It will be also helpful in future corpus research for identification of candidate word combinations as scientific and technical terms.